Learning to express motion events in English and Korean ...
Transcript of Learning to express motion events in English and Korean ...
Cognition. 41 (1991) 83-121
Learning to express motion events in English and Korean: The influence of language-specific lexicalization patterns*
Soonja Choi Department of Linguistics, San Diego State University, 5178 College Ave.. San Diego, CA 92182.
U.S. A.
Melissa Bowerman Max Planck Institute for Psycholinguistics, Wundtlaan 1. 6.c2.5 XD Nijmegen, The Netherlands
Abstract
Choi, S.. and Bowerman. M., 1991. Learning to express motion events in English and Korean: The
influence of language-specific lexicahzation patterns. Cognition. 41: 83-121.
English and Korean differ in how they lexicalize the components of motion events. English characteristically conflates Motion with Manner, Cause, or Deixis, and expresses Path separately. Korean, in contrast, conflates Motion with Path and elements of Figure and Ground in transitive clauses for caused Motion, but conflates motion with Deixis and spells out Path and Manner separately in intransitive clauses for spontaneous motion. Children learning English and Korean show sensitivity to language-specific patterns in the. way they talk about motion from as early as 17-20 months. For example, learners of English quickly general- ize their earliest spatial words - Path particles like up, down, and in - to both spontaneous and caused changes of location and, for up and down, to posture changes, while learners of Korean keep words for spontaneous and caused motion strictly separate and use different words for vertical changes of location and posture changes. These findings challenge the widespread view that children initially map spatial words directly to nonlinguistic spatial concepts, and suggest that they are influenced by the semantic organization of their language virtually from the beginning. We discuss how input and cognition may interact in the early phases of learning to talk about space.
*We would like to thank Herb Clark, Steve Pinker, Dan Slobin, and two anonymous reviewers for helpful comments on earlier drafts of this paper.
OOlO-0277/91/$12.20 0 1991 - Elsevier Science Publishers B.V.
84 S. Choi & M. Bowermun
Introduction
In recent studies of lexical semantics, the expression of motion and location has
played a central role (Jackendoff, 1983, 1990; Levin, 1985; Talmy, 1975, 1985).
Spatial meanings are clearly fundamental to human cognition, and the system for
encoding them is important not only in its own right but also because it provides
the core structuring principles for many meanings that are not fundamentally
spatial.
Although all languages seem to analyze motion/location events into compo-
nents such as Motion and Path, they differ both in how they combine these
notions into words (Talmy, 1975. 1985) and in the categories of spatial relations
they distinguish (Bowerman. 1989, 1991; Casad & Langacker, 1985; Lakoff,
1987). The presence of both universality and language specificity allows us to raise
basic questions about the relationship between nonlinguistic cognition and lan-
guage input in children’s acquisition of spatial expressions. Children are able to
learn a great deal about space on a nonlinguistic basis (Gibson & Spelke, 1983;
Piaget & Inhelder, 1956). But this nonlinguistic knowledge is not enough:
children must still discover how spatial information is organized in their language.
How do these two sources of structure interact in the course of language
development?
Language specificity in semantic organization has rarely been considered in
studies of the acquisition of spatial expressions. Most investigators have assumed
that the meanings of spatial words like in, on, and under reflect nonlinguistic
spatial concepts rather directly. This assumption has been a basis for a major
hypothesis about the acquisition of spatial language: that children learn spatial
terms by mapping them to concepts of space that they have formulated in-
dependently of language (e.g., H. Clark, 1973; Slobin, 1973).
The hypothesis of cognitive priority has found considerable support in research
on the acquisition of spatial words. For example, children acquire English spatial
prepositions and their counterparts in other languages in a relatively consistent
order, and this order seems to reflect primarily the sequence in which the
underlying spatial concepts are mastered (E. Clark, 1973; Johnston & Slobin,
1979). Also, when researchers have compared children’s nonlinguistic grasp of
spatial concepts directly with their knowledge of the words that encode these
meanings, they have invariably found an advantage for nonlinguistic knowledge
(e.g., Corrigan, Halpern, Aviezer, & Goldblatt, 1981; Halpern, Corrigan, &
Aviezer, 1981; Levine & Carey, 1982). These findings have contributed not only
to assumptions about spatial semantic development but also the rise of the more
general “cognition hypothesis”: that children initially identify words, inflections,
and combination patterns with meanings formulated independently of language
(see Cromer, 1974, for discussion). The findings are also consistent with Slobin’s
(1985) proposal that children map grammatical morphemes onto a starting set of
universally shared meanings or “grammaticizable notions”.
Learning to express motion events in English and Korean 85
Nonlinguistic spatial understanding is, then, important in the development of
spatial words. But there is reason to doubt whether, as claimed, it directly
provides spatial concepts to which words can be mapped (see also Van Geert,
1985/6). In examining early vocabularies across languages, Gentner (1982) found
that words for relational meanings are consistently learned later than words for
concrete objects. After ruling out various other explanations (e.g., that adults
model object words more often than relational words), she argued that this
discrepancy reflects differences in the cognitive ‘naturalness” of the correspond-
ing concepts: object concepts are more “given”, whereas relational concepts are
more imposed by the structure of language and so require additional time to be
constructed. Schlesinger (1977) also rejected strong cognitive determinism and
argued on theoretical grounds for an interaction in early language acquisition
between children’s own concepts and the semantic categories modeled in the
input language. And Bowerman (1978a) and Gopnik (1980) proposed that
children at the single-word-utterance stage generalize words to novel referents on
the basis of not only their nonlinguistic concepts but also their observations of
regularities across the situations in which adults use the words.
Existing cross-linguistic studies do not show which is more important in very
young children’s acquisition of spatial language: nonlinguistic spatial knowledge
or the semantic organization of the input language. By age 3, English- and
German-speaking children differ strikingly from Spanish- and Hebrew-speaking
children in how they express spatial information in a story-telling task (Berman &
Slobin, 1987). And 4- to 6-year-old children learning Warlpiri, an Australian
language, differ from children learning English in the meanings they associate
with spatial terms (Bavin, 1990). But it is not clear whether language specificity is
present from the outset, as Gentner (1982) would predict, or emerges only
gradually with divergence from a shared starting point, in line with Slobin (1985).
In this paper, we try to disentangle nonlinguistic spatial cognition from the
structure of the linguistic input by comparing children acquiring English and
Korean. We first contrast the way motion is lexicalized in the two languages, and
then examine spontaneous speech from the period of one-word utterances and
early word combinations.
The lexicalization of motion events in English and Korean
Semantic components of a motion event
In Talmy’s analysis of how languages express motion, a “motion event” is defined
as “a situation containing movement of an entity or maintenance of an entity at a
stationary location” (1985, p. 60). By “movement” is meant a “directed” or
“translative” motion that results in a change of location. By “location” is meant
86 S. Choi & M. Bowwmun
either a static situation or a “contained” motion that results in no overall change
of location (e.g., jumping up and down, walking around in place). In this paper
we focus on movement, along with a limited - although developmentally im-
portant - set of “contained” events, posture changes.
According to Talmy, a (dynamic) motion event has four basic components:
Motion: Presence of motion.
Figure: The moving object.
Ground: The reference-point object with respect to which the Figure moves.
Puth : The course followed by the Figure with respect to the Ground.
These components can be identified in a straightforward way in the following
English sentence:
(1) John went into the room .
[Figure] [Motion] [Path] [Ground]
A motion event can also have a “Manner” or a ‘Cause”, which are analyzed as
distinct external events. To this collection we will add “Deixis” (e.g., motion
towards the speaker vs. away from the speaker), which seems to play a role in the
lexicalization of motion events that is comparable to that of Manner or Cause (see
DeLancey, l985).
According to Talmy, there are fundamental typological differences among
languages in how a motion event is characteristically expressed in a sentence. In
particular, he describes three patterns for what components are expressed by the
main verb root and what is expressed by additional elements. WC will be
concerned with two of these: lexicalization or “conflation” in the main verb of
Motion with either Manner or Cause, with path expressed separately, and
conflation in the main verb of Motion with Path, with Manner or Cause expressed
separately. (In the third, less common pattern, the main verb expresses Motion
plus information about the Figure, with both Path and Manner or Cause
expressed separately.)
English: Conflution of Motion with Manner or Cause
In English, as in most Indo-European languages and Chinese, Motion is charac-
teristically conflated with Manner or Cause, and Path is expressed separately by
‘Talmy (1985, p. 126) apparently regards Deixis (which he terms “Direction”) as closely related to
path in his analysis of the components of a motion event, and Aske (1989) treats it as a kind of Path.
However, Deixis often patterns differently from other kinds of Paths in the way it is lexicalized (e.g.,
many languages express Deixis in main verbs like come and go even though they do not typically
express other kinds of Paths in the verb system). so we distinguish it in this paper.
Learning to express motion events in English and Korean 87
prepositions or particles (Talmy, 1975, 1985). The combination [Motion +
Manner], for example, is found in The rock SLIDIROLLEDIBOUNCED down
(the hill), John WALKEDISKIPPEDIRAN into the room, and phn SLIDI
ROLLEDIBOUNCED the keg into the storeroom. The combination [Motion +
Cause] is seen in The wind BLEW the napkin off the table and John PUSHEDI
THREWIKICKED the keg into the storeroom. The combination [Motion +
Deixis] is also found in English, as in John CAME/ WENT into the room and John
TOOKIBROUGHT the keg into the storeroom.2
As these sentences illustrate, English uses the same verb conflations in both
intransitive sentences expressing spontaneous motions and transitive sentences
expressing motions caused by an agent. In addition, it marks Path in the same
way in sentences of both types, using prepositions and particles like in(to), out
(of 1, on(to), off, UP, d own, and away. It also applies individual Path markers to a
broad range of events within the domains of spontaneous and caused motion. For
example, (put) on is used for the placement of clothing or other items onto all
parts of the body, as well as for actions like putting a cup on a table, a lid on a
jar, and a cap on a pen. Similarly, up and down are used not only for overall
changes in the Figure’s location (e.g., go up, run down) but also with posture
verbs to indicate “in place” changes in the Figure’s alignment or height with
respect to the vertical axis, for example: She suddenly SAT UP (from a lying
posture)lSAT DOWN (from a standing posture); she LAY DOWN (to take a
nap); He STOOD UP (and left the room). (Posture verbs plus up or down also
sometimes express static postures, e.g., he SAT DOWN all during the concert;
see Talmy, 1985).
Korean: Mixed conflation pattern
In the second class of languages in Talmy’s typology, which includes Romance,
Semitic, and Polynesian, Motion is characteristically conflated with Path in the
main verb, and Manner or Cause is expressed separately with an adverbial.
Spanish examples with [Motion + Path] verbs include La botella ENTRY a la
cueva (flotando) “The bottle MOVED-IN to the cave (floating)” and La botella
SALI de la cueva (jlotando) “The bottle MOVED-OUT from the cave
‘All transitive verbs that express caused movement incorporate a causative meaning. This inherent
causativity is distinct from the component “Cause” in Talmy’s analysis. For example, although bring and take are inherently causative, they do not specify an independent event such as kicking or pushing that makes the Figure move, so they are not analyzed as conflations of Motion with Cause. Conversely, although intransitive blow is not inherently causative (cf. the wind blew). it can express
the conflation [Motion + Cause]. as in The napkin blew offthe table (=the napkin moved off the table. from the wind blowing on it). The conflation of Motion with Cause in intransitive sentences is
somewhat restricted in English and will not concern us further; this allows us to use the term “intransitive” to refer to constructions that express spontaneous motion by the Figure.
88 S. Choi & M. Bowermun
(floating)“; compare also SUBIR “move-up”, BAJAR “move-down”, and
PASAR “move-past/through”. Transitive Spanish verbs of the same type include
METER “put in”, PONER “put on”, JUNTAR “put together” and SEPARAR
“take apart” (Talmy, 1985).
Korean presents a mixed picture. In transitive clauses for caused motion, it
conflates Motion with Path, like Spanish. But in intransitive clauses for sponta-
neous motion, it encodes Motion, Path and (optionally) Manner or Cause with
separate constituents, a pattern not described by Talmy. In clauses of both types,
Korean expresses most Path information with verbs; it lacks a system of mor-
phemes dedicated to Path marking like the English spatial prepositions and
particles. However, it does have three locative case endings, EY “at, to”, -LO
“toward”, and -EYSE “from”, which, when suffixed to a Ground nominal,
function like the Spanish prepositions a “to” and de “from” in the examples
above.
The basic word order of Korean is subject-object-verb. The verb phrase
contains one or more “full” verbs; that is, verbs that can occur as the main verb
or alone as a complete utterance (e.g., KA “go”, imperative). The final verb of a
sentence, which may be either a “full” verb or an auxiliary, bears all the
inflectional suffixes such as tense (Lee, l989). A pre-final verb is linked to the
final verb by a “connecting” suffix such as -E or -A; the verbs together form a
so-called compound verb.
Spontaneous motion
In expressions of spontaneous motion, the main (rightmost) verb is usually
KATA “go” or OTA “come”, in which motion is conflated with Deixis.3 This
verb is preceded by a Path verb, which in turn may be preceded by a Manner
verb. The pattern is thus [Manner] [Path] [Motion + Deixis], as illustrated in (2):
(2) John-i pang-ey (ttwui-e) tul-e o-ass-ta.
J.-SUBJ’ room-LOC (run-CONN) enter-CONN come-PST-DECL
[Figure] [Ground] ([Manner]) [Path] [Motion + Deixis]
“John came in(to) the room (running).”
‘This verb can also be TANITA. which means “go and come repeatedly”. However, TANITA
does not combine with all Path verbs, and so is not as productive as the deictic verbs KATA “go” and
OTA “come”.
‘The following abbreviations arc used:
SUBJ - Subject marker
OBJ - Object marker LOC Locative marker
CONN - Connecting suffix
PST - Past tense marker
DECL - Declarative ending CAUS - Causative suffix
Learning to express motion events in English and Korean X9
Table 1. Lexicalization of spontaneous changes of location in Korean
Conflation pattern: [Path] ~ [Motion + Deixis]
(Path] [Motion + Deixis]
OLLA
NAYLYE
TULE
NA
CINA
‘ITALA
THONGHAY
KALOCILLE
TULLE
“ascend”
“descend”
“enter”
“exit”
“pass”
“along”
“through”
“across” “via”
KATA “go”
OTA “come”
Example: OLLA KATA (ascend go) “go up”
In this example, JOHN is the Figure, and PANG “room” is the Ground. The
locative suffix -EY “to, at” on PANG indicates only that PANG represents the
goal or location of the event specified by the verb. The fact that John changed his
location is specified by the rightmost verb, 0- “come”. John’s path with respect to
the Ground is specified by the verb before 0-: TUL- “enter”. John’s Manner of
motion is specified by TTWUI- “run”. Path verbs in addition to TUL- include
NA- “exit”, OLL- “ascend”, and NAYLY- “descend”; a complete list is given in
Table 1.
We translate intransitive Korean Path verbs with words like “enter” and “exit”
instead of “in” and “out” to emphasize that they are verbs. But these translations
are somewhat misleading, since they suggest that the verbs inherently conflate
Motion with Path. In fact, the sense of motion in sentences like (2) comes
primarily from the final verb, KATA “go” or OTA “come”. If a Path verb is
combined with ISSTA “be located” instead, it expresses static location. Path
verbs can also be used as the main (rightmost) verb, in which case they express
motion in a rather abstract, holistic way.’
‘For example, in (1) below, TUL- “enter” is used with ISSTA “be located” to convey the static
situation of the Figure. Similarly, in (2). OLL- “ascend” is the main verb. This sentence conveys the
event of John’s climbing the mountain as a whole; the fact that he had to move is backgrounded and
not central to the meaning. In contrast, the sentence in (3), with the deictic verb KATA “go”, expresses John’s dynamic motion in climbing the mountain.
(1) cui-ka sangca-ey tulle iss-ta.
mouse-SUBJ box-LOC enter-CONN be-DECL
“The mouse is in the box.”
(2) John-i san-ey oil-ass-ta.
John-SUBJ mountain-LOC ascend-PST-DECL “John climbed the mountain.”
(3) John-i san-ey olla ka-ss-ta.
John-SUBJ mountain-LOC ascend go-PST-DECL
“John went up (onto) the mountain.”
As noted earlier, the English Path particles up and down are used to express
not only changes of location (with verbs like go and run), but also changes of
posture (with verbs like sir, stand, and lie). Korean, in contrast, expresses posture
changes with monomorphemic verbs, for example, ANCTA “sit down”, NWUP-
TA “lie down”, (ILE)SETA “stand up”, ILENATA’ “get up”, KKWULTA
“kneel down”, KITAYTA “lean against”. When these posture verbs are pre-
ceded by the Path verbs OLL- “ascend” and NAYLY- “descend”, the resulting
phrase does not have the same meaning as English stand LOP, sit down, etc.: it
specifies that the Figure first gets up onto a higher surface or down onto a lower
surface, and then assumes the indicated posture.
Cuused motion
While spontaneous motion is encoded in “exploded” fashion in Korean, in that
Motion, Path, and Manner are specified by separate words, caused motion is
expressed quite compactly with inherently causative transitive verbs that conflate
[Motion + Path]. Table 2 lists some frequent transitive Path verbs.
Recall that, in English, Path is marked the same way whether a motion is
spontaneous or caused (cf. The ball rolled INTOIOUT OF the box vs. John rolled
the ball INTOIOUT OF the box). For English speakers, it is so natural to think of
these two Path meanings as “the same” that it is hardly worth remarking on. In
Korean, however, Path is usually marked by different forms in the two cases; note
that the only verb roots that appear in both Table 1 (intransitive Path verbs for
spontaneous motion) and Table 2 (transitive Path verbs for caused motion) are
OLL- “ascend” and NAYLY- “descend”.’ (These roots are inherently intransi-
tive; the transitive forms are derived by adding the causative suffix -1.)
Not only are Path forms different for spontaneous and caused motion, but so
are most of their meanings. Consider notions of joining and separation (bringing
an object into or out of contact with another), which are typically expressed in
English with phrases like put inlonltogether and take outlofflupart. These are
encoded in Korean with a variety of verbs, as shown in Table 2.” KKITA
(glossable loosely as “fit”, but used much more widely than English fir) is
indifferent to whether the Figure goes into, onto, over, or together with the
Ground, as long as it leads to a tight fit/three-dimensional meshing; hence, it is
routinely used to express putting earplugs INTO ears and a cassette INTO a
cassette case, one Lego piece ONTO or TOGETHER with another, and the top
“The morpheme ILE may have the meaning “ascend” but it occurs only with these two posture
verbs. optionally with SETA “stand up”. and obligatorily in ILENATA “get up”. ‘The two verbs TUL- “enter” and NA “exit” can also take the causative suffix. However. their
causative forms are not productive spatial verbs because they cannot stand alone, and when they
combine with other verbs they have idiomatic senses. e.g.. SON-UL NA-I MILTA “hand-OBJ
exit-CAUS push” (=put hand out to shake hands or receive something). “If a Ground nominal is included in the sentence. it is marked with the suffix -EY “at. to” or
-EYSE “from”. as appropriate.
Learning to express motion events in English and Korean 91
Table 2. Korean transitive verbs for caused motion
Verb Meaning (Examples)
Cause to ascendldescend
OLLITA Cause something to ascend.
(Move a poster upward on the wall)
Cause something to descend.
(Move a poster downward on the wail) NAYLITA
Joiniseparate
KKITA
I PPAYTA
NEHTA
I KKENAYTA
PWUTHITA
/TTEYTA
KKOCTA
TAMTA
IKKENAYTA
SIT-I-A
PWUSTA
IPHWUTA
NOHTA
KKATA
KKAKTA
“Fit”/“unfit” one three-dimensional object to/from another.
(Leg0 pieces, ear plugs-ears. cassette-cassette case, top-pen.
ring-finger)
Put/take things in/out of a loose container.
(wallet-handbag, ball-box. furniture-room)
Join/separate a flat surface to/from another flat surface.
(sticker-book, poster-wall, two table sides)
Put a solid object elongated in one dimension into/onto a base.
(hower-vase, book-shelf, dart-board, hairpin-hair)
/Separation: PPAYTA when the base holds the figure tightly,
KKENAYTA when it holds it loosely
Put/take multiple objects in/out of a container that one can carry.
(fruits-basket, candies-bowl, toys-box)
Load something into/onto a vehicle.
(hay-truck. package-car, car-boat)
/Separation: NALUTA (“move an object from one place to another”)
when the object is moved to another place, but NOHTA (see below)
when the object is put down on the ground
Pour liquid (or a large quantity of tiny objects) into/out of a container.
(milk-cup, sand-pail)
Put something loosely on a surface.
(pen-table, chair-floor)
Separation: TULTA for focusing on taking the object into the hand.
CIPTA for focusing on picking it up.
Take off a covering layer or wrapper.
(shell-nuts, peel-banana, wrapper-candy)
/Joining: SSATA for wrapping an object tightly.
Take off a covering layer with knife.
(skin-apple, planing a board, sharpening a pencil)
Put clothing item onto one’s own body part
IPTA Trunk of body (dress, shirt. pants)
SSUTA Head (hat, umbrella)
SINTA Feet, legs (socks, shoes)
CHATA Waist, wrist (belt, watch, diaper)
(PESTA is the reverse of all of these)
Put something onto/into one’s own body part in order to support or carry it
ANTA Arms (a person, an object, e.g., baby, package) EPTA Back (a person. e.g., a baby or child on mother’s back) CITA Back (an object, if not also supported by shoulder) MEYTA Shoulder (an object hanging, e.g., backpack. bag over shoulder) ITA Head (an object, e.g.. a pot) TULTA Hand (an object, e.g., briefcase, suitcase) MWULTA Mouth (an object, e.g., a cigarette)
Y2 5’. <'hoi & M. Bowermm
of a pen ONTO (=“OVER”) the pen. The reversal of these actions is specified
by PPAYTA “unfit”. Because of the KKITAiPPAYTA Path category, Korean
speakers must distinguish actions of putting in/on/together that result in a fitting re-
lationship (KKITA) from those that result in loose containment (NEHTA) or sur-
face contact (NOHTA, PWUTHITA); similarly, they must distinguish “taking out”
of tight versus loose containment (PPAYTA vs. KKENAYTA) and “taking off”
or removal from attached versus loose surface contact (PPAYTA or mEYTA
VS. CIPTA). These groupings and distinctions in Path meanings are not made in
expressions for spontaneous movements into or out of a container, onto or off a
surface. etc., since KKITA and PPAYTA do not have intransitive counterparts.
Transitive Path verbs of joining and separation also contrast with intransitive
Path verbs in that they incorporate aspects of Figure and Ground as well as Path.
For example, different verbs are used for solid versus liquid Figures, for three-
dimensional versus flat versus elongated Figures, and for Ground objects that are
conventionally used for carrying things versus vehicles versus other kinds of
Grounds (see Table 2). Fine distinctions are made when the Ground is part of the
human body: there are different verbs for putting clothing onto different body
parts, and also for putting people or objects into/onto the arms, back, shoulder,
head. mouth, and hand for purposes of support or carrying. Acts of putting a
Figure onto the back are distinguished according to whether the Figure is animate
or inanimate.
In Korean, expressions for caused motion also differ from those for sponta-
ncous motion with respect to Deixis. Recall that intransitive expressions of
spontaneous motion typically have as main verbs KATA “go” or OTA “come”.
which conflate Motion with Deixis. But for caused motion, Korean has no deictic
transitive verbs comparable to English take and bring. Self-initiated changes of
location by animate beings are consistently encoded with intransitive deictic
verbs. When someone “takes” or “brings” something while moving, this can be
expressed by combining KACY-E “have” with KATA “go” or OTA “come”; for
example, John tooklbrought u book to the librury is rendered as JOHN-l
CHAYK-UL TOSEKWAN-EY KACY-E KA-IO-ASS-TA “John-SUBJ book-
OBJ library-LOC have-CONN go-/come-PST-DECL” (=John went/came to the
library having a book).
In transitive clauses, just as in intransitive clauses, Manner can be expressed
with a verb preceding the Path verb, for example, TOLLY-E PPAYTA (turn
“unfit”): “take Figure from its tightly fitting ground by turning it; twist out/off/
apart”. The pre-final verb can also express Cause, for example, MIL-E NEHTA
(push put-in): “put something in a container by pushing it; push in”.” However,
“Certain Path-conflating transitive verbs can also be used as Manner verbs in combination with a second Path-conflating transitive verb. because they express some Manner information in addition to
Path information. For example. KKITA “fit” and PPAYTA “unfit” suggest that the action requires a
bit of force; hence, one can say KKI-E NEHTA (fit put-in) to express shoving a block of a certain shape through a matching hole in a child’s shape-fitting box so that the block falls down inside.
Learning to express motion events in English und Korean 93
these combinations are less frequent than constructions like twistipulllcutlroll off
and pushlthrowlkicklslide in in English. This is because the two languages differ
in what information must be expressed and what can be left to inference.
In English it is often obligatory to spell out Path rather completely, even when
it can be readily inferred from context. If we heard “John threw his keys TO his
desk/TO the drawer”, we could reasonably suppose that the keys ended up ON
the desk or IN the drawer. Still, these sentences sound odd: oy1 and in arc needed.
Even when it is grammatical to specify Path less completely, fuller information is
often given, especially in everyday speech; compare John took his keys FROM his
deskIFROM the drawer (a bit formal or bookish) with John took his keys OFF his
desklOUT of the drawer (completely colloquial). In Korean, in contrast, a Path
verb can often be omitted if a transitive verb expressing the Manner or Cause of
the motion is supplied. As long as the Ground is specified and the relationship
between Figure and Ground can be easily inferred, locative case endings such as
-EY “to, at” or -EYSE “from” on the Ground nominal arc sufficient, and a Path
verb often sounds redundant.“’
Summary of the lexicalization of motion events in English and Korean
To summarize, English uses the same verb conflation patterns in both intransitive’
clauses expressing spontaneous motion and transitive clauses expressing caused
motion, and it encodes Path separately with the same Path markers (particles and
prepositions) whether the clause is transitive or intransitive. Korean, in contrast.
uses different lexicalization patterns for spontaneous motion and caused motion,
and most of its Path markers (verbs) in the two cases are distinct. An overview of
these patterns is given in Table 3. In addition, many Korean Path verbs have a
“‘For example, in (1) below, the Ground “desk” is specified with the locative marker (-ey), and
the main verb is the Cause verb. In (2), both Cause (TENCY-) and Path (NOH-) verbs are present in
addition to the Ground; to Korean speakers this seems redundant.
(1) John-i yelswey-lul chayksang-ey TENCY-ess-ta.
John-SUBJ key-OBJ desk-LOC throw-PST-DECL
“John threw keys TO desk.”
(2) ? John-i yelswey-lul chayksang-ey TENCY-e NOH-ass-ta.
John-SUBJ key-OBJ desk-LOC throw-CONN put-on-PST-DECL
“John threw keys ONTO desk.”
Fuller Path information can be supplied in sentences like (I) by a finer specification of the Ground object; for example:
(3)
(4)
John-i yelswey-lul chayksang-wui-ey TENCY-ess-ta.
John-SUBJ key-OBJ desk-top-LOC throw-PST-DECL
“John threw keys TO desktop.”
John-i yelswey-lul selhap-an-ey TENCY-ess-ta. John-SUBJ key-OBJ drawer-inside-LOC throw-PST-DECL
“John threw keys TO inside of drawer.”
‘$4 S. Choi & h4. Bowerman
Table 3. Comparison of English and Korean conflation patterns for motion events
English Korean
Verb
[Motion + Manner]
[Motion + Cause]
[Motion + Deixis]
Spontaneous motion
Particle Verb
[Path] [Manner]
Causer1 motion
Verb
[Path]
Verb
[Motion + D&is]
Verb
[Motion + Manner]
[Motion + Cause]
[Motion + Deixis]
Particle
[Path]
Verb
[Manner]
[Cause]
Verb
[Motion + Path + Ground]”
“These verbs are inherently transitive and causative
narrower range of uses than the English Path markers that translate them. For
example, OLL- “ascend” and NAYLY- “descend” are used only for changes of
location and not posture, while English uses up and down for both. Similarly,
Korean uses different verbs for putting clothing onto different parts of the body
and for placing objects onto other surfaces, while English uses on across this
whole range.”
“In addition to verbs conforming to the characteristic conflation pattern of English described by
Talmy (1975, 1085). English has a number of intransitive and transitive Path-confating verbs. Many of
these, including enfer, exir. ascend. descend, insert, extruct, join. and separate, are borrowings from
Romance. In Romance they represent the basic pattern. whereas in English they belong to a more
formal register than their native counterparts go in/our/up/down, put inifogether, take outiaparf. A
few path verbs, such as fall. rise, and raise, are native to English. Notions of Motion and Path and
sometimes Figure or Ground also seem to lurk in the more complex meanings of a variety of other
verbs such as pluck, stuff. jam. pee/, load, fit. and unwrap. In light of such verbs. Steve Pinker
(personal communication) has suggested to us that the differences between English and Korean might
be “more in the number and frequency of verbs used than in some major typological parameter of the
entire language”. But we believe that the differences are more fundamantal. In Korean, Path
meanings are expressed almost exclusively by Path verbs (only “at/to” and “from” are expressed
separately. and only if the Ground object is mentioned). In English, however, most native Path verbs
may or even must combine with a separate preposition or particle that either marks the incorporated
Path meaning redundantly or specifies it more precisely; compare full DOWN. rise/raise UP, srujyljam X INTO Y, pluck X OFFIOUT of Y, peel X OFF YIX and Y ASART. load X ONTOIINTO Y. f;t X INTO/ONTO/TOGETHER wirh Y. This is true even of some of the Romance borrowings, e.g..
insert X INTO Y, not simply _. TO Y. Candidate Path verbs may be absorbed into the basic English
pattern of marking Path separately because they often incorporate elements of Manner as well as Path and so can be treated as [Motion + Manner] conflations; for example fall means something like “to go
down in an uncontrolled manner”, insert suggests “to put in in a controlled way“ (e.g.. because the
space is small). stufy suggests the use of force. and pee/ specities a particular manner in which two
surfaces separate.
Learning to express motion events in Engli.vh and Korean 95
Development of motion expressions in English and Korean
Children learning English and Korean must, then, master different systems for
lexicalizing motion events. How do they approach this task? The English data we
use to investigate this question come from Bowerman’s diary records of her two
daughters, C and E. Data collection began when the children were about 1 year
old and cover the period of one-word utterances and early word combination in
rich detail. Aspects of the expression of motion in these records have been
discussed in Bowerman (1976, 1978a, 1978b, 1980). The data can be compared
with information given in a number of studies of children learning English,
including Bloom (1973), Farwell (1977), Gopnik (1980), Greenfield and Smith
( 1976), McCune-Nicohch ( 1981)) and Tomasello ( 1957).
Our main set of Korean data was collected longitudinally by Choi, who visited
four children in their homes every three to four weeks from age 14 to 24-28
months (group I). At each session, she and the mother played with the child for
60-90 minutes. All sessions were video-taped and transcribed. Choi also elicited
mothers’ reports on their children’s uses of spatial expressions. These data are
supplemented by data collected from four other Korean children every two to
four weeks from 19-20 months to 25-34 months (group II).”
Early in language development, most references to action take place in the
immediate context of the action. In this study we consider only utterances*
produced while a motion event was taking place, just after it had occurred, or just
before it occurred as a statement of intention, desire, or expectation. Both the
English- and Korean-speaking children began to use words to encode motion in
such situations in the same age range - around 14 -16 months.
The motion events referred to by the two sets of children were remarkably
similar. For example, they commented on their own changes of posture or
location, such as sitting down, standing up, or climbing up onto chairs or laps;
they appealed to adults for help in changing location or to go outside; they asked
to be picked up or carried; and they referred to donning and doffing clothing and
to object manipulations of many kinds, for example, putting things into a bag and
taking them out and putting Lego pieces or Popbeads together and taking them
apart. Some examples are shown in Table 4. These similar preoccupations - also
shown by Dromi’s (1987) Hebrew-speaking daughter at the one-word stage - are
apparently driven by shared aspects of children’s general cognitive development,
including what they are interested in talking about (Gopnik & Choi, 1990;
Gopnik & Meltzoff, 1986).
Underlying the impression of similarity between the two sets of children there
are important differences. We look first at the children learning English and then
“Pat Clancy collected the data from two of these children, and Young-Joo Kim from one; the
fourth was studied by Choi. We would like to thank Clancy and Kim for their generous permission to use their data.
96 S. Choi & M. Bowermun
Table 4. Words produced in similar contexts by learners of English and Korean between 14 and 21 months
Context
Wanting to go outside
Asking M to pick her up
Sitting down
Asking M to get up
in the morning
English Korean“
out pakk-ey “outside-LOC”
UP anta “pick up and hold in arms”
down ancta “sit down”
up ilenata “get up”
Joining two Lego picccs
Separating Popbeads
Putting coat on Putting toys in container
Putting a small object
into a hole or a crack
on
Off
on
in
in
kkita 3”
ppayta “unfit”
ipta “put clothes on trunk”
nehta “put in loosely”
kkita ‘Tit”
“The verb ending -TA on all but the first example is the citation form. Endings
actually produced by the children include various medals like -E (or -A) for requests or
assertions (e.g., un-a for un-fu “pick up and hold in arms” and am-a for am-tu ‘sit
down”). and -TA for certain types of assertions (Choi. 1991).
at the children learning Korean, focusing on spontaneous versus caused motion,
motion “up” and “down”. and how Path morphemes combined with verbs.”
English
Words used for motion events It is well known that words like down, up, in, out, on, off, back, and away play
a central role in the early expression of motion by children learning English, first
appearing as single-word utterances and later figuring in early word combinations
(Bloom, 1973; Farwell, 1977; Gopnik, 1980, Greenfield & Smith, 1976; Gruendel,
1977a; McCune-Nicolich, 1981; Miller & Ervin, 1964; Tomasello, 1987). This is
true also for our two diary subjects, C and E (Bowerman, 1976, 1978a, 1980). In
adult speech. these words often appear as verb particles in sentence-final position
with heavy stress, which may make them especially salient to children (Brown,
1973; Slobin, 1973; Tomasello, 1987). Many of them also serve as prepositions in
adult English, and can express static location, as in The book is IN the bookcase. However, children at first use them primarily or exclusively for motion. C and E
began to use them for static location in the second half of the second year; for
“See Bowerman (1989) and Bowerman and Choi (in preparation) on the expression of caused
motion. especially how children categorize manipulations like putting things into/onto/together with other things and taking them out/off/apart.
Learning to express motion events in English and Korean 97
example, in produced while peeking into the bag in which a hamburger had just
arrived in a restaurant (E, 19 months). Other words that C and E used for motion
events between 14 and 18 months include go, come, sit, walk, run, jump, ride, faii, push, pull, and throw.
Spontaneous versus caused motion Table 5 shows the emergence of Path particles and verbs for spontaneous
versus caused motion in C’s and E’s speech. Utterances are classified according to
Table 5. Early words for spontaneous and caused motion events in C’s and E’s speecha
c: Age in
months Soontaneous motion Caused motion
14-16 downb. out
17-18 up, down, out, on
come. fall, walk. run, sit, ride
19-20 up, down, in, out, away
go. come, walk. fall, sit, lie
go away
21-22 up, down, in, out, back, away
23-24
go, come, walk, jump, fall, sit, open, close. push, turn, pour, ride carry, come, put, take
come down, sit down. get down,
fall down. lie down, get/got up,
sit up, come out, get/got out,
fall out, sit on, come back, go away
fall down, he down, help down,
sit up, get up, go in, pour in.
get out, pour out, keep out, come o come out, fall out, take off,
clean off. dry off, put back,
throw away, take away, put away,
get away
up, down, in, out, on, back
go, come, run, jump, fall,
sit, ride
fall down, sit down, he down,
get down, get/got up_ get in,
go out. come out, fall off,
come back, go away
on’, off’
opend
up, down, on, off
open
up, down, in, out, on, off, back
open, close, hold, fall
up, down, in. out, on. off,
back, away
up, down, in, out, on, off. away
open, close, push, throw, put,
take, give, spill, pour
get down, push down, get up,
pick up. pull up, bring in, do in,
take out, pour out, blow out
(=deflate), put on, take off, brush off, came off, take away,
put away. throw away, put together
98 S. Choi & M. Bowerman
Table 5 continued.
E:
Age in
months Spontaneous motion Caused motion
14-16
17718
up. down. out
go, come
out’
go, walk. jump, fall. sit
lY_20 up. down. in. out, on, off. back.
away
go, come, walk. run. fall, ride
21-22
come up. get up, stand up, step up,
sit down, lie down. fall down.
run down, come in, come out.
get out, going on, come/came off.
get off, come back, get away
stand up. get down. pour in,
close in. came on. came off,
take off, fall off. get off.
throw away
up. down. in, out. on. off. back. up. down, in. out, on. off, back. away away
go. come. run. walk. fall. climb.
ride
open. close, push, pull, take.
put. bring. give. turn, kick,
carry. fall. spill, pour
get up. stand up. came up. carry up, get up. pull up, put
reach up. play up, go up and down, down. push down, pull down,
come down, fall down, get down. pour down. put in. get in. push in,
lie down, come in, get in, sit in. fit in, pour in, dip in. take
come out, fall out, get out, out, pull out. get out. carry
stick out. blow out (i.e.. go out of out, put on. get on, take on.
window), get on, come/came off. take off. get off. push off.
come/came back, going away, came off, fell off. put back.
going around give back. throw away
down, off, back
open, close, push, pull, throw
Up. on. off. in, back
open. close, push, pull. throw, sit, fall
up. down. in. out. on, off, back,
away
open, close, push. pull. spill.
pour, kick, throw. take. come
carry. fit
“Particles and verbs are listed only if they were produced spontaneously (i.c.. not
imitated) at least three times during the period shown. either in isolation or in
combination with other words. All non-imitated verb + particle combinations are listed.
hMost uses of down in this period were for getting off a rocking horse. sometimes
with an adult‘s help. One instance was for going downstairs.
‘Until 18 months C pronounced both on and off as /a/ (final consonants in general
were missing), so it is unclear whether she had two words or one. “Although open and close are not used in adult English for caused motion (i.e..
change of location). the children often overextended them to contexts in which adults
would say take off/au//apart or put on/in/together (see Bowerman. 197Xa). ‘E often said up and clown for static position in this period (e.g.. up while looking at
something high on a shelf.)
Learning to express motion event.s in English and Korean 99
whether the motion was (or would be, in the case of anticipated events)
spontaneous, or required an agent’s causal action. Most utterances classified
under “spontaneous motion” were used for self-initiated motion by animate
beings, usually the child herself (see Huttenlocher, Smiley, & Charney, 1983).
They also included uses of up and down for changes of location that the child
initiated and was active in pursuing, even though she was helped by an adult, for
example, up as the child clambered up on a chair with a boost. (Up or down as
requests to be picked up or lifted down, or comments on these actions, were
classified as caused motion.)
Utterances were classified under “caused motion” when they referred to a
motion brought about by an external agent. When a child says “in” while putting
a ball into a box, we cannot be certain whether she intends to refer to the agent’s
action (“put in”), or only to the motion of the Figure (“go in”). Adult English
often allows the speaker to focus only on the Figure’s motion, leaving the agent
out of perspective (Talmy, 198.5), and children in the early period of word
combining say both “put X in” and “X goes in” (for example) in the context of
caused motion (Bloom, Lightbown, & Hood, 1975). Our classification thus uses
the nonlinguistic context as a guide - imperfect in the case of caused motion - to
the child’s likely intentions.
In the age period 14-16 months, C and E produced only a few Path particles.
All but one (E’s down) were applied to either spontaneous motion or caused
motion, but not both. In some cases this reflected the child’s initial restriction of
the form to specific contexts; for example, at first C said out only for going
outdoors (spontaneous) and E said off only for removing clothing and other
objects from the body (caused). In other cases, however, the child used the form
quite productively within the limits of spontaneous or caused motion (see
Bowerman, 1978a, 1980, on C’s extensive use of oy1 and off).
Over the next few months, however, the children used Path particles increas-
ingly often for both spontaneous and caused motion. By 19-20 months, they used
almost all Path particles in both ways, and for a wide variety of spontaneous and
caused motion events. For example, they used on for sitting or standing on things
and for putting on clothing of all kinds, attaching tops to pens and stickers to
surfaces, and putting objects down on surfaces, and off for the reverse of these
actions. They used in for going into houses, rooms, bathtubs, and the child seat of
a shopping cart and for putting things into various containers (e.g., pieces into
puzzles, noodles into bowl, riding toys into house), and out for the reverse of
these actions (see Gopnik, 1980, and Gopnik & Meltzoff, 1986, for similar uses by
other children). They used back for their own or another person’s spontaneous
return to an original location, for putting objects back where they were usually
kept (e.g., watch on arm, books on shelf), and for rejoining parts of an object
(e.g., top on pen, lid on bottle). Between 17 and 20 months they also used many
100 S. Choi & M. Bowerman
of the particles for static spatial relations, for example, in when looking at a box
with crackers in it or a picture of a bear in a helicopter.
Combining Path with Manner1 Cause1 Deictic verbs
By 19 months (E) and 21 months (C), the children began to combine Path
particles with a variety of verbs specifying the Manner. Cause, or Deictic aspect
of a motion event (see Table 5) (combinations with nouns naming the Figure
started earlier). Many of the children’s verb-particle combinations are also
common in adult speech, but there is evidence that they understood the underly-
ing combinatorial principle and were not simply imitating. First, they produced
novel combinations such as “carry up” (picking up and righting a fallen-over
stool; E, 21 months), “sit in” (after another child got into a bus and sat down; E,
21 months), “close in” (trying to stuff jack-in-the-box down into box and shut lid;
E, 20 months), “catch in” (asking M to capture her between two boxes; E, 24
months), “do it in” (=put it in; C, 23 months), and “blow out” (a) holding hand
out of open car window; E, 22 months; (b) asking F to deflate a beach ball; C, 24
months). Second, the particle and the verb in the children’s combinations factored
motion events appropriately into an independent Path and Motion; for example,
out expressed a Figure’s exit from containment regardless of whether the action
was specified as fall, pour, or take, while the use of pull,‘push, fall, etc., was
indifferent to whether the Path followed by the Figure was specified as up, down,
in, out, on, off, or back.
DOWN and UP
We will illustrate English-speaking children’s use of Path morphemes more
closely with down and up. These are typically among the first words used for
motion events: one or both sometimes appear as early as 12-14 months (e.g.,
Farwell, 1977, Greenfield and Smith, 1976; Gruendel, 1977a; Nelson, 1974), and
they are often present by 16 or 17 months (Bloom, 1973; Gopnik, 1980; Ingram,
1971; Tomasello. 1987). In E’s speech, down appeared at 13 months and up at 16
months; in C’s it was 16 and 17 months. Both children occasionally overextended
down to “up” situations before learning up, an error also reported by Greenfield
and Smith. In Table 6, we show representative uses in chronological order for
each child.
Like many children reported in the literature, C and E at first said up and/or
down primarily or exclusively for movements of their own bodies, either sponta-
neous (including assisted) or caused by an adult. But they soon became more
flexible. Between 16 and 20 months, both children said up and down for their own
and other people’s spontaneous vertical motions, including both changes of
location like falling and getting on or climbing off raised surfaces such as chairs,
couches, riding toys, and laps, and changes of posture like sitting down, standing
Learning to express motion events in English and Korean 101
Table 6. Examples of DOWN and UP in C’s and E’s early speech
Spontaneous motion Caused motion
(age in months) DOWN (age in months)
c: Wanting M to help her get down from rocking horse (16)
Climbing down from doll’s Pushing cat’s head down (17)
crib (17) Taking cow down out of crib (17)
Sliding down off bed (18) Wanting M to take her out of
Rot-n-Spin chair (18)
Coming downstairs (18)
Climbing down out of washtub (19)
Watching squirrel come down
tree (20)
Coming downstairs (21)
Wanting M to take her down from
dressing table (19)
E: Trying to climb down off counter (13)
At top of slide wanting
to slide down (14)
Wanting C to come down from counter (15)
Asking M to sit down (16)
Sitting down in car (16)
Climbing down from chair (16)
Asking F to sit down (17)
After getting from chair to floor (19)
Getting down from high chair (20)
c: Climbing up steps (17)
Wanting to get on upper bunk bed (18)
Wanting to climb on counter (18)
Climbing onto the couch
(18) Wanting to get into M’s
tap (18) Wanting M to stand up
by the crib (19) Wanting F to get out of
bed (19)
Wanting M to take C down from
Asking M to take her down from
chair (13)
counter (14)
Dumping an armload of yarn into
her wagon (16)
Setting books on the floor (16)
Wanting to take chair down from
on table (16)
Wanting M to put beads down on
the floor (17)
Wanting M to put her cup down
on saucer (18)
Wanting M to take cup down from
desk (20)
UP
Trying to get her walker up onto
the couch (18)
102 S. Choi & M. Bowrrmun
Table 6 continued.
Spontaneous motion Caused motion
(age in months) (age in months)
c:
UP (continued)
Wanting M to get out of Picking up crayons from the
bed (20) floor (21)
When somebody picks up a hahy
(21) Picking up a piggy hank and
taking it to a pile of
toys she’s making (21)
E: Standing up in the car (16)
Climbing up the slide (16)
Standing up in her crib (16)
Climbing up on her horse (16)
Trying to lift herself up on
the counter (16) Re: M who just climbed up on
a chair (17)
Sitting up after lying on her
hack (19) Climbing up on a chair (19)
Trying to climb on M’s
lap (20) Standing up in high chair
(20) When C arrives at top of
stairs (21)
Putting a tiny figure on a
toy tree (17)
Putting something on coffee
table (1Y)
Putting a peg doll on top of
toy fire engine (IY)
Picking up a newspaper (20)
Wanting Mary to lift her onto
a bed (20) Wanting M to pick her up (21)
No/c: utterances were produced just before, during. or just after the events indicated.
M = Mother. F = Father, C = Child’s sister.
UP, and getting up in the morning. They also said up and down for caused
motions, for example, when they wanted an adult to lift them up onto a higher
surface or take them down from it, for picking up objects from the floor or
putting them on raised surfaces, pushing or pulling things downward, and putting
things down on the floor or other low surfaces. They also used up as a request to
be picked up and held or carried, and both up and down for static situations, for
example, up when pointing to the upper branches of a tree in a picture, and down
when looking at a doll floating head down in the tub. This range of uses is
consistent with that reported for other children in this age period.
English-speaking children acquire up and down so early, and extend them so
readily to many situations of vertical motion or orientation, that many inves-
tigators have assumed that vertical motion *‘up” and “down” are nonlinguistic
notions. For example, Nelson (1974), reporting on a year-old child who extended
up on the first day of its use “to all vertical movement of the child himself or of
objects”, proposed that “there is a core representation of this action concept .
Leurnmg to express motion events in English and Korean 103
something like Vertical Movement” (p. 281). Similarly, Bloom (1973, p. 70)
suggested that the early uses of up reflect the “notion of ‘upness”’ and Gruendel
(1977a) concluded that uses of up in her data support Bloom’s proposal that
“‘upness’ is itself a true early-cognized or conceptualized relation”. In a study of
the development of relational words at the one-word stage, McCune-Nicolich
(1981) found that up and down, along with several other words, emerged
somewhat abruptly in the speech of five children, spread rapidly to a variety of
contexts, and were less likely to be imitated than other words. She proposed that
these early-learned relational words code “pre-established cognitive categories” -
in particular, operative knowledge of the late sensorimotor period to do with
space, causality, sequence, and the like. She predicted that “since operative
intelligence is a universal aspect of cognition, the same categories of meaning
would be expected for all children, although various lexical items might be used to
encode these” (p. 18).
When our attention is confined to English, it is plausible to think that children
generate notions of vertical motion nonlinguistically and simply map them directly
to up and down. But in cross-linguistic perspective, it is sobering to realize how
neatly tuned these meanings are to the requirements of what is, after all, a
language-specific system of expressing Path. Let us turn now to Korean to see
whether children exposed to a different system express the same meanings, albeit
mapped to different words.
Korean
Words used to express motion Because we have fewer data from each Korean child than from our English-
speaking subjects, we will often consider the children of a particular age period
together. A summary of pooled data is presented in Table 7.
Like our English-speaking subjects, our Korean subjects began to refer to
motion events between 14 and 16 months. The first productive words for motion
of all four children in group I were the transitive Path verbs KKITA “fit” and/or
PPAYTA “unfit”.‘” Typical contexts of use included putting Lego pieces together
or taking them apart, and fitting plastic shapes into the holes of a shape box. By
17-18 months a number of other transitive Path verbs emerged: PWUTHITA
“put one surface to another”, which the children used for stickers and bandaids,
KKA(K)TA “peel off”,” NEHTA “put into a loose container”, KKENAYTA
“take out of loose container”, and some “carrying” and “clothing” verbs. By
“The ending -TA marks the citation form of a verb in Korean. Verbs in the children’s speech were
suffixed instead with various modal endings. most typically -E (or -A), which is used in adult speech for requests or assertions. Thus, a child’s rendering of KKITA “fit” would typically be KKI-E.
“At this early stage of development, the children did not differentiate phonologically between KKATA “take off covering layer or wrapper” and KKAKTA “take off covering layer with knife”.
104 S. Choi & M. Bowerman
Table 7. Early words for spontaneous and caused motion events in Korean children’s speecha
Age in Spontaneous motion Caused motion
months
14-16 O.M.:h kkita (I) “fit”. ppayta (3) “““tit”.
(N = 4) yelta (I) “open”. tatta (I) “CIose”
17-IH DEIXlS: kata (1) “go” 0 M kklta (4) “fit”. ppayta (4) “unlit”.
(N = 4) yelta (I) “open”. tatta (I) “CIUSC”. pwuthtta (I) “,uxtapose two surfaces”,
POSTURE: ancta (2) “sit down” kka(k)ta (3) “peel off”. nehta (I) “put I”“.
kkenayta (I) “take out”.
“lllta (I) “cause to go up”
CARRYING: anta (2) “I” arms”, epta (I) “on back”
CLOTHING. pesta (1) “take off”
MANNER/CAUSE: tolhta (I) “turn”
19-20 DEIXIS: kata (3) “go”, ota (I) “come”
(,V = 4)
POSTURE: ancta (2) “ut down”,
ilenata (3) “get up”
PATH t DEIXIS: “a kata ( I) “20 out”
MANNER: ttwuita (I) “run
0.M : kklta (4) “fit”. ppayta (4) “““tit”.
yelta (3) ““pen”. tatta (I) “close“.
pwuthita (I) “juxtapose twosurfaces”.
kka(k)ta (3) “peel off”. nehta (I) “put I”“.
nohta (I) “put on surface”.
olhta ( I ) -cause l” go up”.
kkocta (I) “put elongated object to base”.
kkenayta (I) “take out”
CARRYING: anta (2) “I” arms”. epta (2) “on back”
CLOTHING, pest” (I) “take off”,
,pta ( I) “on trunk”.
sinta (I ) “0” feet”
MANNER/CAUSE. toll0 (Ii “turn”
21-22 DEIXIS: kata (4) “go”. eta (4) “come’
(n;=S)
POSTURE: ancta (4) “sit down”.
ilenata (3) “get up”,
nwupta (2) “lie down”.
ilebeta (I) “stand up”
PATH t DEIXIS: na kata/ota (4) “go/come out”.
tule katalota (l)“go/comem”.
olla kata!ota (3) “go/come up”,
naylye kata!ota (I) “go/come down”.
ttele kata (I) “fall-go”
MANNER: ketta (I) “walk”,
ttwuita (I) “run”
naluta (I) “fly”
MANNER + DEIXIS: “alla kata (I) “fly-g””
(change of location by flymg)
0.M : kkita (6) “fit”. pfxdyta (6) “unfit”,
yelta (3) “open”, tatta (2) “close“.
pwthita (3) “]uxtaposc two surface\“.
kka(k)ta (7) “peel off”. nehta (3) “put in”.
kkenayta (3) “take out”.
ollita (3) “CBUSC to go up”.
naylita (1) “cause to go down“.
kkocta (I) “put elongated object to base”
CARRYING: anta (5) “in arms”, epta (5) “on hack”
CLOTHING: pest” (4) “take off”.
lpta (4) “on trunk”. sinta (4) “on feet
ssuta (3) “on head”
MANNER:CAUSE: hsotta (I) “pour carele\aly”.
chata (I) “kick”, milta (3) “push”.
nwuluta (I) “push down”.
tolhta (I) “turn”. tencita (1) “throw”
Table 7 continued.
Learning to express motion events in English and Koreun 105
Age in Spontaneous motion Caused motmn
months
23-24 DEIXIS: kata (7) “go”. ota (5) “come” O.M.: kkita (6) “fit”. ppayta (6) “unfit”.
(N=X) yelta (7) “open”. tatta (3) “close”.
POSTURE: ancta (7) “sit down”. pwuthita (4) “luxtapose two surfaces”.
denata (4) “get up”. kka(k)ta (7) “peel off”. nehta (4) “put in”,
nwupta (3) “lie down”. kkenayta (3) “take out”.
ileseta (1) “stand up” ollita (2) “cause to go up”,
nayhta (2) “cause to go down”.
PATH t DEIXIS: na k&iota (5) “go/come out”,
tule kataiota (4) “go/come III”,
olla kataiota (5) “go/come up”,
naylye kataiota (2) “go/come down”
MANNER. ttwuita (2) “run”.
ttuta (l)%at”.
ttelecita (I) “fall“
nohta (3) “put on surface”.
kkocta (I) “put elongated object to base”
CARRYING: anta (6) “m arms”. cpta (4) “on back”,
tulta (I) “III hands”
CLOTHING. pesta (6) “take ofl”.
ipta (5) *‘on trunk”. smta (4) “on feet”,
ssuta (2) “on head”
MANNER/CAUSE. tollita (I) “turn”,
nwuluta (2) “push down”,
tenclta (2) “throu”.
kkulta ( I) “pull”.
capta (2) “holdicatch”
“Numbers m parentheses refer to the number of children who produced the word durmg the period shown. (N =I) refers to group 1.
(N = 8) refers to groups I and II combined. For group I, each verb listed was produced by the child at least once during the recording
scss~on, and the mother reported that the child produced it more than once during the age period Indicated. For group II. the herb was
produced by the chdd at least once during the recording session.
hO.M. = verbs of object mampulation.
19-20 months, NOHTA “put on surface”, KKOCTA “put elongated object to
base”, and OLLITA “cause to go up” were added, along with additional clothing
verbs.
Spontaneous versus caused motion
As Table 7 shows, the Korean children used transitive Path verbs only for
caused motion and never overgeneralized them to spontaneous motion; for
example, they never said KKITA “fit” when they crept into a narrow space or
KKENAYTA “take out of a loose container” when they got out of the bathtub.
In fact they never violated the distinction between spontaneous and caused
motion along a Path throughout the entire developmental period observed: no
verb was used in contexts of both kinds. In comparison, recall that our English-
speaking subjects used some Path particles for both spontaneous and caused
motion by as early as 14-16 months, and many by 20 months.lh A major
‘“The English-speaking children did discriminate well between transitive and intransitive verbs.
They never used a transitive verb such as take for spontaneous motion. Occasionally they used
intransitive verbs for caused motion (see Table 5), but mostly in contexts where this is also acceptable
in adult speech, e.g.. fair when a Figure is dropped or knocked over and come (out/off. etc.) for manipulations of small objects (cf. adult utterances like “Will it come out?“). Errors such as “I come
(=bring) it closer” (Bowerman, 1974) did not start until about age 2. and can be attributed to a learned rule, not ignorance that a verb is basically intransitive (Bowerman, 1974; Pinker. 1989).
difference between children learning English and Korean, then. is in their
willingness to extend Path words across the transitivity boundary.
Unlike English-speaking children. our Korean subjects at first focused almost
exclusively on caused motion. In C’s and E’s speech, expressions for spontaneous
and caused motion developed in parallel. In the Korean children’s speech.
intransitive motion verbs appeared much later than transitive motion verbs. As
Table 7 shows, the children produced no intransitive verbs for Motion or Path at
all between 14 and 16 months. At 17-18 months KATA “go” was used by one
child. Two others may have also said KATA at this age according to their
mothers, but, if so, it was far less productive than transitive motion verbs: the
children did not say it during the recording sessions and the mothers’ reports were
not consistent from one session to the next. KATA “go” and OTA “come”
became productive only at 19 months. One child combined KATA with the Path
verb NA “exit” (NA KATA “exit go; go out”) during this period as a request to
go outside. The other children began to combine KATA and OTA with Path
verbs only at 21 months.
As discussed earlier, many transitive motion verbs of Korean conflate Motion
not only with Path but also with information about Ground and sometimes
Figure. The children’s use of these verbs was generally appropriate, showing that
they were sensitive to the incorporation of these elements. The sense that ground
may be a component of a motion verb’s meaning seemed to become particularly
strong between 17 and 20 months. At this time the children distinguished two
verbs of supporting/carrying according to the body part that serves as Ground
(ANTA “put into arms to support/carry” vs. EPTA “put on back to support/
carry”), and they also began to distinguish clothing verbs according to the
Ground body part: IPTA “put clothes on trunk”. SINTA “put clothing (e.g.,
shoes. socks) on feet”. SSUTA “put clothing on head” appeared at 21 months.
Combining Path verbs with Manner and Cause verbs
Our Korean subjects were rather slow to learn verbs like TENCITA “throw”
and MILTA “push”, which in adult speech can be used either alone or in
combination with transitive Path verbs to express the Manner or Cause of a
caused motion event. Only one such verb. TOLLITA “cause to turn”, is attested
up through 20 months, and it was produced by only one child. More verbs of this
type began to come in at 21-22 months (see Table 7). But they were not
combined with Path verbs. even though the children produced word combinations
of other kinds. Caused motion events were expressed either with Path verbs or
with Manner/Cause verbs, but not with both at once - a pattern characteristic of
adult Korean as well, as discussed earlier.
Our Korean subjects contrast sharply with our English-speaking subjects in
their slow acquisition of Manner/Cause verbs and their failure to combine them
with Path verbs. Recall that in the age range 17-20 months, C and E expressed
motion events with both Path particles and many different Manner/Cause verbs,
Learning to express motion events in English untl Korem I07
and from 19-21 months they often combined the two elements, particularly when
expressing caused motion (see Table 5). Such combinations are. of course,
characteristic of English and other languages of its conflation type, as described
by Talmy (1975, 19SS).
Motion “down” and “up” As discussed earlier, English-speaking children learn down and up so early,
and extend them so readily to many events involving downward and upward
motion, that many investigators have supposed that they are mapped directly to
nonlinguistic sensorimotor notions of vertical motion “downward” and “upward”.
If this is so, Korean children-presumably equipped with similar nonlinguistic
concepts - should seize on Korean words produced frequently in contexts involv-
ing vertical motion, and extend them freely to other events involving vertical
motion regardless of whether the motion is spontaneous or caused or whether it
involves a change of location or posture. For example, they might initially say
either OLLA KATA “go up” or OLLITA “cause to go up” for both spontaneous
upward motions, including posture changes, and for caused upward motions.
Similarly. they might say either NAYLYE KATA “go down” or NAYLITA
“cause to go down” for getting down. sitting or lying down, putting things down,
and as requests to be put down. Alternatively, they might select ANTA “pick up
and support/carry in arms” to mean “up” in general. or ANCTA “sit down” to
mean “down” in general.
This does not occur. Although ANTA “pick up and hold/carry in arms” and
ANCTA “sit down”‘were produced by some of the children from 17-18 months.
they were never overextended to other situations involving vertical motion. The
intransitive and transitive causative forms of OLL- “up” and NAYLY- “down”
emerged very late compared to up and down in the speech of children learning
English. The development of our four youngest subjects (group 1) is shown in
Table 8.
Among the children of group I, SN was, at 18 months, the youngest to produce
one of these words - OLLITA “cause to go up”. However, he made an intriguing
error in the meaning he first assigned to it. When he was 17 months old his
mother had said OLLITA when she was putting plates back in a kitchen cabinet
high above the counter. SN apparently overlooked the “up” information embed-
ded in this complex event and inferred that the word meant “put something in the
location where it belongs”; for a month, he used the verb only for “putting away”
events of many sorts, regardless of directionality, for example, putting a toy back
in a container on the floor. He made this error at a time when he was learning a
variety of transitive verbs - for example, clothing verbs - that include a Ground
component. The acquisition of Ground-incorporating verbs may sensitize Korean
children to the possibility that Ground may be relevant to the meaning of a new
transitive motion verb. Only at 19 months did SN begin to use OLLITA for
caused upward motion, for example, to ask his mother to lift him up onto a step.
108 S. (‘hoi & M. Bowerrnan
Table 8. Exumples of NAYLY- (“descend”) and OLL- (“ascend”) in four Korean-speaking children
Spontaneous motion
(age in months)
Caused motion
(age in months)
AN:
NAYLYE KATA (descend-go)
Getting down from her high chair
(24)
NAYLY-ITA (descend-causative)
Taking a plate down from table (24)
Getting down from counter (24)
MK: Taking an object down from counter
(25) Taking an object down from shelf
(25)
SN: _
YN: Asking M to pull her pants down
(23) Asking M to pull her pants down
(2.5)
Getting down from her high chair
(26)
Asking M to take her down from her
high chair (25)
Going downstairs (26)
Getting down from a step (26)
AN:
OLLA KATA (ascend-go)
Climbing up on couch (24)
Climbing onto her bed (24)
Climbing up in her high chair
(24) MK: -
OLL-ITA (ascenddcausative)
Putting an object up on the table (21)
Putting a toy on her leg while seated
(22) Putting her plate up on the counter
(24)
Putting an object on the counter (26)
Putting a toy car up on the shelf (27)
SN: Putting toys back in their usual place
(1X) (not necessarily “up”, see text)
Wanting M to put him in high chair
(19) Wanting M to lift him onto the step
in the bathroom (19)
Going upstairs (22)
Putting an object on the chair/
kitchen counter (20)
YN:
Climbing up on a chair (22)
Climbing up in her high chair
(26)
Asking M to pull her pants up (26)
Asking M to lift her up onto a
stool (26)
Learning to express motion events in English and Korean 109
At 22 months he finally also began to say OLLA KATA “go up” in connection
with spontaneous motions like getting on a chair.
OLLITA and OLLA KATA emerged even later in the speech of the other
three children. The development of NAYLITA “cause to go down” and NAYLY-
E KATA “go down” is similar to that of OLLITA “cause to go up” and OLLA
KATA “go up”, but still slower, as shown in Table 8. Although the Korean
children were slow to use words comparable to up and down, this does not mean
that they did not talk about events involving upward and downward motion. They
did - but using verbs that classify these events on the basis of criteria other than
their shared Path.
The late appearance of intransitive Path verbs is not restricted to OLLA
KATA “go up” and NAYLYE KATA “go ‘down”, nor to these four youngest
children of our sample. Of the eight children whose data at 21-22 months are
shown in Table 7, only three said OLLA KATAiOTA “go/come up”, only one
said NAYLYE KATA “go/come down”, only four said NA KATAiOTA
“go/come out”, and only one said TULE KATAiOTA “go/come in”. We return
to the question of why these verbs are so late in the discussion section.
Discussion
Although children learning English and Korean talk about similar motion events
in the second year of life, they do not do so in similar ways. English-speaking
children rely heavily on Path particles. They start out using some of these in
restricted or idiosyncratic ways, but soon extend them to a wide range of
spontaneous and caused motion events that share similar abstract Paths. By about
20 months they begin to combine them productively with verbs that specify the
Manner, Cause, or Deictic aspects of the motion event.
Korean children use no words in these ways. Like Korean adults, they
distinguish strictly between words for spontaneous and caused motion. Concen-
trating first on caused motion, they learn a variety of transitive verbs that conflate
Path with notions of Figure and especially Ground, and extend them to different
classes of motion events than are picked out by English-speaking children’s Path
particles.” Their intransitive Path verbs are limited for many months to verbs of
“For example, they use the same verb (KKITA “fit”) for putting a Figure into a tight container
and attaching it to an outside surface (in vs. on for learners of English), and the same verb (PPAYTA
“unfit”) for the reverse of these actions (out vs. off). But they use different verbs for putting objects
into tight versus loose containers (KKITA vs. NEHTA; both in for learners of English) or taking them out (PPAYTA vs. KKENAYTA; both out), for joining three-dimensional (KKITA), flat
(PWUTHITA), OK elongated (KKOCTA) Figures to a Ground (all in or on for learners of English.
depending on whether there is containment), for putting clothing on the head (SSUTA), trunk
(IPTA), or feet (SINTA; all on), and for being supported or carried in the arms (ANTA) or on the back (EPTA) (both up or carry).
posture change. They do not acquire intransitive Path verbs for spontaneous
motion “in”. “out”, “up”, and “down” until long after English learners begin to
use Path particles for spontaneous motion, and they are just as late on transitive
verbs for caused motion “up” and “down”. Once they do learn verbs for “up”
and “down”, they never overgeneralize them to posture changes or use them as
requests to be picked up and carried, both favorite uses of up and down by
English-speaking children.
These findings challenge the widespread view that children map spatial mor-
phemes directly to their sensorimotor concepts of space, and suggest instead that
children are guided in constructing spatial semantic categories by the language
they are exposed to. We will elaborate on this interpretation shortly. But first let
us try to rule out alternative interpretations that do not require crediting such
young children with a language-specific semantic organization of space.
Context-bound learning and homonyms
Perhaps the look of language-specific semantic organization is an illusion. Maybe
children just imitate the words they hear in particular contexts, and see no
relationship between them. For instance, our subjects may have simply learned
what to say when climbing onto a chair, when wanting to be picked up, and when
getting into a standing posture. Learners of Korean would use three different
verbs, while learners of English would say up in each case, but for both sets of
children the word said in each context would be independent.
This hypothesis is easy to rule out. First, our subjects used spatial words
creatively, extending them to many events for which they had never heard adults
use them. Many of their novel uses were completely appropriate; for example,
l-l 1 and 17-24 in Table 9. Others were errors from the adult point of view, for
example, 12-16 and 25-34. Errors show particularly clearly that children are not
simply imitating what they have heard in particular contexts. They have often
been interpretated as evidence that children rely on meanings generated in-
dependently of language (e.g., Nelson, 1974; see Bowerman, 1989, for
discussion). But our subjects’ errors seem to reflect problems of detail within
spatial semantic systems that, in broad outline. were already language specific.
For example, our Korean subjects knew that PPAYTA “unfit” had to do with
taking something from a position of tight fit or attachment, but they sometimes
overextended it to attachments of the wrong kind, for example. those involving
flat surfaces (e.g., 32 in Table 9) or tight clothing or embrace (e.g., 25-28, 31).
Similarly, our English-speaking subjects knew that in applied generally to “con-
tainment”, but they tended to assimilate “position between” to this category
(14-15)‘”
‘“kc also Bowerman (lY78a, 1980) on overextensions of opn and close to actions that adults would encode with on and o# or rog~~hrr and uparr.
Lraming to expms tnotwn wents in English unri Korrcrn III
Table 9. Examples of novel uses of spatial words by learners of English and Korean between 15 and 25 months (age in months; errors are starred)
ENGLISH
1. ON 2. ON + negative
head shake.
3. OFF
4. IN. TOY IN
5. IN ‘GAIN
6. OUT
7. OUT
X. SMOKE OUT
9. DOWN
10. DOWN
11. UP. DOWN
12. *OPEN
13. *CLOSE KNEES
14. *MONIES IN
IS. *IN
16. *OFF
KOREAN
17. PPAYTA
18. PPAYTA
19. PPAYTA
20. PPAYTA
21. PPAYTA
22. KKITA
23. KKITA
24. KKITA
25. *PPAYTA
26. *PPAYTA
27. *PPAYTA
2X. *PPAYTA
20. “PPAYTA
30. *KKITA
31. *KKITA
32. *KKITA
33. *OLLITA
34. *KKOCTA
Putting a ladybug magnet on a can opener, C In.
Has just been told not to pull off a bit of paper stuck to
M‘s leg. hut she wants it off, E 1X.
Asking M to remove a (nonremovable) upright pole from back
of her riding toy, C 17.
Trying to tit piece of camera into loop formed by pull-handle
of drawer. C 21.
Trying to shove piece of toy furniture through door in doll
house, E 18.
Has just dipped hand into her glass of milk and taken it out
again; is now inspecting it, C 17.
Trapped hehind toys in her room . she wants help in getting
out, E 17.
Watching steam coming out of vent in the ground, E 21.
Pushing down head of neighbor’s cat, C 17.
Asking M to move chair from tahle to Hoor. E 16.
“Walking” her fingers up to her neck and back down. E IY.
Trying to separate two Frisbees. C 16.
Asking M to put her knees together. E 21.
Looking for coins she’d just stuffed down between two couch
cushions, E 1Y.
Putting ping-pong ball hetwccn knees. E 20.
Asking M to unfold a newspaper. C IS.
Trying to pull out the string from the end of the toy fire hose.
AN 15.
Trying to take out Investigator’s (Inv.‘s) jigsaw puzzle game
from tight-fitting box. AN IS.
Asking Inv. to take lid off her (Inv.‘s) pill box. SN IY.
Taking Hute apart. HS 22.
Trying to take out pencil stuck through paper, HS 22.
Putting doll into tight-fitting seat of small horse, AN 17.
Fitting a train into its wooden base (Inv.‘s new toy). MK 17.
Watching Inv. put video cassette in camcorder. TJ 23.
Trying to take bib/shirt off, AN lo. (PESTA is appropriate.)
Asking M to take his bib/shirt off. SN IX. (PESTA)
Asking M to take her shirt off. TJ IX. (PESTA)
Wanting to get toy away from sister. HS 22. (CWUTA “give”).
Asking someone to peel a banana, HS 25. (KKATA)
Sticking fork into apple, TJ 23. (KKOCTA)
Re: Being held tight by an adult. PL 22. (ANTA)
Attaching magnetic fish to magnetic mouth of duck. TJ 25.
(PWUTHITA)
Putting toys hack in place, SN IX. (KACTA TWUTA “bring/ take hack”)
Putting a Lego piece onto another. SN 20. (KKITA)
Second, the “context-bound” explanation flies in the face of much work on
early word use by other investigators. Although many researchers have noted that
first words are often tied to specific contexts, most assume that this phase is
short-lived. According to some, there is a shift to a more symbolic basis for word
meanings around the middle of the second year (e.g., McShane, 1979; Nelson &
Lucariello, 1985); others argue that many or most words are never significantly
context bound at all (Barrett, Harris, & Chasin. 1991; Harris, Barrett, Jones, &
Brookes, 1988; Huttenlocher & Smiley, 1987). The move away from context-
based word use is often assumed to show that the child has come to rely on her
own nonlinguistic conceptualizations of objects and events (e.g., Barrett et al.,
1991; Nelson & Lucariello, 1985). Our subjects began to use a variety of spatial
words in flexible and context-free ways between 16 and 20 months. Ironically,
though, this development went paired with striking language specificity, which
clashes with the hypothesis that the children were now starting to rely on their
own nonlinguistic conceptions of space.
But perhaps we can reconcile evidence for creativity with the idea that children
learn words for rather specific meanings. Suppose children share a repertoire of
nonlinguistic spatial concepts that, although narrow, are broad enough to accom-
modate new instances. And suppose they associate each concept with a different
word. For the hypothetical concepts “sitting down”, “lying down”, “going
down”, and ‘*putting down”, Korean children would learn four different words.
while English speakers would learn four words all pronounced down, perhaps as
reduced versions of more complete English verb phrases like sit down, lie down.
go down, and put down. Let us call this the “homonym” hypothesis.‘”
This hypothesis requires a close look. Even for adult English speakers, some
uses of the same Path particles are probably unrelated, and other uses only
loosely related via a network of polysemes (see Brugman, 1981; Herskovits, 1986;
Lakoff, 1987; Langacker, 1987; Lindner, 1981). And some uses that are related
for adults might well start out as distinct for children, only coming together later
as learners discover abstract similarities across situations to which the same
particles are applied. We ourselves have assumed that English-speaking children’s
uses of on and o@ for actions with light switches and water faucets are in-
dependent of the spatial uses of these morphemes, and so have left them out of
our analysts. But most of the differences we have found between children
learning English and Korean do not submit easily to the homonym hypothesis.
The hypothesis requires us to assume that whenever learners of one language
use a single word for situations that learners of another language distinguish with
two or more words, the single word actually has two or more independent
meanings. Sometimes this seems plausible; for example, it is not too jarring to
“‘We are grateful to Eve Clark for making us worry about this possibility, and for her insightful
feedback on our attempts to deal with it.
Learning to express motion events in English and Korean 113
posit homonymous downs for “sitting down” and “lying down”. But do we really
want to have to separate down for climbing down from a chair from down for
being lifted down from the chair, and in for climbing into a tub from in for putting
something into the tub? (Recall that the Korean children used different path
verbs for spontaneous and caused motion events.) And do we feel comfortable
with homonymous ins for putting a book in a tight-fitting box versus a looser box
(KKITA vs. NEHTA for Korean children), and homonymous outs for the reverse
of these actions (PPAYTA vs. KKENAYTA)?‘” These uses are so consistent with
the central spatial meanings of these particles that it has never occurred to
previous investigators that they might be independent acquisitions.
In fact there are good reasons to believe they are not. Once particular spatial
words emerge in children’s speech they often spread rapidly to new uses, which
supports the intuition that they are interrelated. For example, our subject C first
said in at 19 months for coming indoors (=“come/go in”). Within a few days she
also used it for “put in” actions like putting a sock in the laundry basket and a
bead in a container, and for static containment, as when playing with an
unopened thermometer package. E first said up at 16 months when she stood up
in the car (=“stand up”); within a few days she also used it for “go/get up”
events like climbing up a slide, stepping up on a little chair, and trying to pull
herself up by the kitchen counter, as a request to be lifted (=“pick up”), and for
static “upness”, for example, for a picture of a cat sitting on a broomstick at a
higher angle than a witch. Similar rapid extension patterns for up have been
reported by Bloom (1973, p. SS), Leopold (1939), and Nelson (1974). Children’s
speed in generalizing especially up and down across diverse contexts is well
recognized, and has often been cited to support the hypothesis that these words
are mapped to unitary concepts of vertical motion (e.g., McCune-Nicholich,
1981). While we disagree that the route between nonlinguistic concepts and
spatial word meanings is as direct as this, we concur that core uses (though not
necessarily every use) of the various Path particles are related for the child.
Language-specific semantic learning
The differences we have found between learners of English and Korean cannot be
ascribed to word meanings that are highly context bound or based on very narrow
nonlinguistic spatial concepts. They constitute real differences in the children’s
“‘Note also that adult English routinely applies put in and take out to most “tight” and “loose”
manipulations with containers, so learners of English are probably not relying on distinct underlying
verb phrases when they say in or out for these acts. If English speakers find it strange to split up the core meanings of in and our, Korean speakers find it equally strange to imagine that when a child says
KKITA for fitting a figure “into” or “onto” a tight ground, or PPAYTA for taking it “out of” or “off”, these uses are independent for him.
semantic organization - differences that correspond directly to the way spatial
meanings are structured in the language the children are learning.
English isolates Path as a recurrent component of motion events in an
exceptionally clear and consistent way. With its system of Path particles, it
encourages learners to identify abstract notions of Path that are indifferent to
whether the Figure moves spontaneously or is caused to move, and to details
about the shape or identity of the Figure and Ground objects. Korean does not
single out Path as a separate component of motion events as clearly and
consistently as English. It uses Path verbs that differ in both form and meaning
for spontaneous and caused motion (except for motion “up” and “down”),
and - for caused motion and posture verbs - it combines information about Path
with information about the shape or identity of the Figure and Ground objects.
Korean children, then, are not prompted to analyze out Path as an abstract
component of motion events as strongly as are learners of English, and this may
account for their delay in acquiring those Path verbs that do express Path in
relatively pure form. Instead, they are encouraged to classify motion events on
the basis of Path meanings admixed with causativity and properties of the Figure
and Ground.”
In rejecting the hypothesis that children’s early spatial words are mapped to
nonlinguistic concepts, we do not mean that nonlinguistic spatial cognition plays
no role in spatial semantic development. Clearly it does. For example, across
languages. children learn words for topological relationships (e.g., on and in)
before words for projective relationships (e.g., in front of and behind) (Johnston
& Slobin, 1979). This bias - also shown by our subjects - presumably reflects the
order of emergence of nonlinguistic spatial understanding. Children also make
certain errors even on words for topological relationships (see Table 9). which
suggests that some topological distinctions are more difficult than others, presum-
ably for cognitive reasons. We must, then, posit an interaction between language
input and cognitive development. But how does this interaction take place?
More than thirty years ago Brown proposed that for language learners, “a
speech invariance is a signal to form some hypothesis about the corresponding
invariance of referent” (1958, p. 228). Our findings confirm this. Even very young
children must be sensitive to the way adults use spatial words across contexts -
otherwise they could not learn to classify spatial referents in a language-specific
way so early. But we still know little about how children track uses of the same
form over time, and how they generate and modify hypotheses about the adult’s
intended “invariance of referent”.
“These claims are intended to apply only to children’s organization of spaceforpurposes oftulking
ahour it (“thinking for speaking”, to borrow a phrase from Slobin, 1991). We take no stand here on
whether the proposed semantic differences between learners of English and Korean have any Whorfian consequences for nonlinguistic spatial thought. Such effects would be compatible with our
findings, but are not entailed by them.
Learning to express motion events in Etlgiish and Korean 115
A prerequisite for generating hypotheses about spatial words is to have some
system for representing space. But “both the nature of the initial system for
internally describing space and the way in which such a system can be modified by
experience . remain as mysterious as ever” (Pylyshyn, 1977. p. 174). Many
researchers have approached this problem by positing a set of semantic primitives
or “privileged notions”: an inborn mental vocabulary of distinctions or compo-
nents drawn on in acquiring spatial words, such as verticality, region, inclusion,
support, contact, attachment, Figure, Ground, Path or direction, and object
dimensionality (point. line, plane, or volume) (Bierwisch, 1967; H. Clark, 1973;
Jackendoff, this issue; Jackendoff & Landau, 1991; Miller & Johnson-Laird, 1976;
Olson & Bialystok, 1983; Talmy, 1983). Spatial primitives would no doubt
interact with other privileged notions such as, for verbs, causality and manner
(Gropen et al., this volume; Jackendoff. 1983, 1990; Pinker, 1989). The reper-
toire of primitives would be the same for all languages, although they might be
combined in different ways.
This approach has a number of advantages. Most important for us, it can help
explain how children home in so quickly on language-specific spatial meanings:
they do not need to generate endless hypotheses about what aspects of spatial
relationships might be relevant in their local language, but only to choose and
combine primitives in the right way. The approach also allows us to reconcile
language specificity with errors: by hypothesis, children are relatively accurate on
words based on (combinations of) features that are highly accessible, but make
errors on words with features that are less salient or emerge only later in cognitive
development (Bowerman, 1985; Slobin, 1985).” But it will take serious work to
make the semantic primitives approach truly explanatory rather than simply
programmatic.
One problem is that it may be difficult to make principled distinctions between
meanings that are “privileged” for space (or any other semantic domain) and
other conceptual distinctions a speaker can make (see also Bolinger, 1965; E.
Clark. 1983). For instance, notions like “verticality”, “inclusion”, and “support”
make plausible-sounding spatial primitives, whereas notions like “arms”, “head”,
“back”, “feet”, and “clothing item” do not. But notions of both kinds played a
role in the meaning of our subjects’ earliest spatial words.
One attempt to get a better grip on what components should be considered
“privileged” has been to restrict the notion of primitives to the “grammaticized”
portion of language. According to this approach, open-class words like nouns and
verbs may incorporate any kind of meaning, no matter how idiosyncratic or
culturally specific. In contrast, closed-class morphemes like inflections, particles,
“Of course. this explanation is circular unless we can find some independent way to predict how
accessible a feature is. See Bowerman and Gentner (in preparation) for a cross-linguistic (Dutch- English) test of the prediction that the case of spatial semantic distinctions for children is linked to the frequency with which the distinctions are marked in the world’s languages.
116 S. Choi & M. Bowerman
and prepositions draw on a much more restricted set of meanings (Slobin, 1985;
Talmy. 1983, 1985). In particular, closed-class spatial morphemes are insensitive
to most properties of the Figure and Ground objects, such as exact shape, angle,
and size, and instead schematize spatial situations in terms of more abstract
topological properties (Talmy, 1983; see also Jackendoff & Landau, 1991).” If
information of this kind comes built in, English learners will not waste time
hypothesizing that closed-class items like up or on apply only to Figures or
Grounds of a certain kind (Landau & Stecker, in press). Korean learners,
however, cannot rule out this possibility when they meet open-class items like the
verb SINTA “put a clothing item on the feet or legs”.
But this solution raises new problems. First, if meanings like “clothing item”
and “feet” are not semantic primitives but have to be constructed from ex-
perience, it should take longer for Korean children to acquire spatial verbs that
incorporate them than for English-speaking children to acquire Path particles that
do not. But this is not the case. Second, we would need to explain how children
between about 17 and 20 months - the period when our subjects were acquiring
language-specific meanings for spatial morphemes - determine whether a mor-
pheme is a member of an open or a closed class, and so decide on what kinds of
hypotheses they should consider.
Finally, the semantic primitives approach probably underestimates what the
child must learn about meaning. When meaning components are assumed to be
built in, there is no need to explain them further (except, of course, at the genetic
level). For some candidate primitives, this may be correct. For example, both
English and Korean learners seem to recognize that different expressions may be
needed for spontaneous and caused motion, and they do not extend words across
this boundary unless - as for English particles - this use is demonstrated in the
input. This is consistent with reports that children learning Japanese (Rispoli,
1987) and Quiche Mayan (Pye, 1985) identify verbs as transitive or intransitive
from the start, and suggests that a full-blown sensitivity to caused versus
spontaneous (or to transitive vs. intransitive) may be present in children from the
outset of language acquisition. For other candidate primitives, however, ex-
perience may have significant “sharpening” effects (along lines discussed by
Bornstein, 1979, for perceptual features). Let us consider the notion “Path” as an
example.
We assume that learners of English and Korean have the same prelinguistic
potential for identifying Path as an independent component of motion events. But
“Levinson (1991) has challenged this argument with data from Tzeltal. In this Mayan language.
spatial relationships comparable to in and on are expressed with a closed-class set of “positional”
verbs that predicate “to be located” of Figures of different types. Far from being abstract and purely
topological. these verbs distinguish Figures on the basis of shape, size, and in some cases identity: for example. different verbs are needed for predicating spatial location of a wide-mouthed vessel, a
narrow-mouthed vessel, an inverted object with flat side down (e.g.. a lump of dough), a small sphere,
a large sphere. things sitting bulging in a bag, objects leaning at various angles. and SO on.
Learning to express motion events in English and Korean 117
we have suggested that the structure of English encourages children to develop
this potential more than the structure of Korean - to actually carry out this kind
of analysis. Why should we think this? Why not simply assume that both sets of
children have a fully developed notion of Path from the beginning, along with
some candidate instantiations of it like motion “up” and “down”? Our reason for
doubt is that our Korean subjects were so late to acquire “pure” Path markers of
Korean like the intransitive verbs OLL- “ascend”, NAYL- “descend”, TUL-
“enter”, NA “exit”, and the transitive verbs OLLITA and NAYLITA “cause to
ascend I descend”. They began to use these verbs only several months after
acquiring verbs in which Path is conflated with information about the Figure
and/or the Ground, and months after our English learners had acquired words
like up, down, in, and out. This delay is hard to explain if we assume that Korean and English learners
both begin with a fully developed ability to isolate Path from complex motion
events.” However, it is interpretable if we assume that children do not have a
fully developed notion of Path, but rather are selectively prompted by the
structure of the input to develop their skill at this analysis. Children learning
English are systematically shown how to isolate a few recurring kinds of Path, and
they learn how to do this quickly. Children learning Korean, in contrast, meet
Path mostly conflated with notions of spontaneous or caused motion and often
with specific properties of the Figure and Ground as well, so it takes them longer
to realize that Path can sometimes be extracted and receive its own marking. If
this analysis is correct, a danger of the “semantic primitives” approach is that by
supplying the child with components that are “ready to go”, it may cause us to
overlook subtle learning processes promoted by the structure of the language
being acquired.
In conclusion, we have shown that the meanings of children’s early spatial
words are language specific. This means that language learners do not map spatial
“‘One attempt to do so would be to say that Korean learners are just as sensitive as English learners
to path, but that they have more trouble with the superficial problem of identifying the morphemes to
express it with (we are following the logic of Slobin. 1973, here). In particular, English Path particles often occur sentence-finally and can receive heavy stress, both properties known to facilitate
acquisition of a form. In contrast. the Korean intransitive “pure” path verbs are usually pre-final,
followed by KATA “go” or OTA “come”. This account does not go through, however, Most
critically. it does not explain why the transitive “pure” path verbs OLLITA and NAYLITA “cause to
ascend/descend” are just as delayed as the intransitive “pure” path verbs. even though they are not
followed by deictic verbs and are identical in their positioning to KKITA “tit” and other transitive
path verbs that are learned much earlier. It is also doubtful whether the intransitive “pure” path verbs
are perceptually less salient than early-learned transitive path verbs. In caregivers’ speech, verbs of both kinds receive major stress and are typically followed by further unstressed morphemes such as the modal markers CWE “do something for someone” or PWA “try”. For example, OLL- “ascend” might appear in a phrase like [olaka) (OLL-A KA) “ascend go; go up” or [olakapwa (OLL-A KA PWA) “try to go up”. and KKITA “tit” in a phrase like [kkiepwa] (KKI-E PWA) “try to put it in” or [kkiecwe] (KKI-E CWE) “shall I put it in for you’?” In both cases, the path verbs are stressed but nominal. which makes them similar in perceptual salience.
11x S. Choi 22 M. Bowcwnun
words directly onto nonlinguistic spatial concepts, as has often been proposed, but
instead are sensitive to the semantic structure of the input language virtually from
the beginning. How children figure out language-specific spatial categories re-
mains a puzzle. Although an appeal to semantic primitives offers some help, it
leaves many questions unanswered. One thing seems clear, however: children
could not learn language-specific spatial meanings as quickly as they do unless
they have some good ideas about what to look for.
References
Aske, J. (1YXY). Path predicates in English and Spanish: A closer look. Berkdc~y Lingui.sfic~.s Socirfy
PupUY, 1.5. l-13.
Barrett. M.. Harris. M.. & Chain. J. (lYY1). Early lexical development and matcrnal speech: A
comparison of children’s initial and subsequent uses of words. Jourr~al 0.f C‘hiltl Lunguugr. 18,
21-30.
Bavin. E. (IYYO). Locative terms and Warlpiri acquisition. Journd of Child Lrrng~ugc’. 17. 43-66.
Berman, R., & Slobin, D.I. (lYX7). Five ways of learning how to talk about events: A crosslinguistic
study of children’s narratives. Berkeley Cognitive Scirnce Reporr. 46: Berkeley. CA.
Bierwisch. M. ( lY67). Some semantic universals of German adjectivala. Fowdution.s of Lung~uge. .7.
l-36.
Bloom. L. (lY73). One word uf II rime. The Hague: Mouton.
Bloom. L.. Lightbown, P.. B Hood. L. (lY75). Structure and variation in child language. Monogruph.>
of‘ rhr Socie/y for Reseurch m Child Developmenr, 40 (serial no. 160).
Bolinger. D. (lY65). The atomization of meaning. L~in~urrge. -II, 555-573.
Bornstein, M.H. (lY7Y). Perceptual development: Stability and change in feature perception. In M.H.
Bornstein & W. Kessen (Eds.). P.sychologicul tlrvelopmcwt from infancy: lrnugr to infention.
Hillsdale. NJ: Erlbaum.
Bowerman, M. (lY7-1). Learning the structure of causative verbs: A study in the relationship of
cognitive. semantic. and syntactic development. Pupers utrd Reports on Child Lunguuge
Del~elopmenr. 8. Id?- 178.
Bowerman. M. (lY76). Semantic factors in the acquisition of rules for word USC and sentence
construction. In D.M. Morehead & A.E. Morehead (Ed>.). Nonnul untl defic,ient child
Iurrguuge. Baltimore: University Park Pres,.
Bowerman. M. (197%). The acquisition of word meaning: An investigation into some current
conHicts. In N. Water\on & C. Snow (Eds.). 771c developmenr of’communicution. New York:
Wiley. Bowerman, M. (lY78b). Systematizing semantic knowledge: Changes over time in the child’s
organization of word meaning. Child Development. 49, Y77-Y87.
Bowerman. M. (1980). The structure and origin of semantic categorics in the language-learning child. In M.L. Foster & S.H. Brandes (Eds.). S~mhol as sense. New York: Academic Press.
Bowerman. M. (1985). What shapes children‘s grammars? In D.I. Slobin (Ed.). The cross-lirlguisric
study of Iungw~ge ucyuisition. Vol. 2: Theoretical xwes. Hillsdale, NJ: Erlbaum.
Bowerman, M. (lY8Y). Learning a semantic system: What role do cognitive predispositions play‘? In M.L. Rice & R.L. Schiefelhusch (Eds.), The truchuhilify of lunguuge. Baltimore: Brooks.
Bowerman. M. (1YYl). The origins of children’s spatial semantic categories: Cognitive vs. linguistic
determinants. Paper presented at Wenner-Gren symposium “Rethinking linguistic relativity”,
Ocho Rios, Jamaica. (To appear in J.J. Gumperz & S.C. Levinson (Eds.). Rethinking linguistic
relariviy .)
Bowerman, M.. & Choi. S. (in preparation). Bowerman, M.. & Gentner. D. (in preparation)
Leurning to express motion events in English und Korean II’)
Brown, R. (195X). Words and things. New York: Free Press. Brown. R. (1973). A first lunguage: The early stuges. Cambridge, MA: Harvard University Press.
Brugman. C. (1981). Story of over. M.A. thesis. University of California at Berkeley. Available from
the Indiana University Linguistics Club. Casad, E., & Langacker. R. (1985). “Inside” and “outside” in Cora grammar. Internutwna/ Journul
of Americun Lingulstic.r. 51. 217-281.
Choi. S. (1991). Early acquisition of epistemic meanings in Korean: A study of sentence-ending
suffixes in the spontaneous speech of three children. First Languuge, 11, 93-l 19.
Clark. E.V. (1973). Non-linguistic strategies and the acquisition of word meanings. Cognition, 2.
161-1X2.
Clark. E.V. (1983). Meanings and concepts. In J.H. Flavell 8: E.M. Markman (Eds.), Mussen
handbook of child psychology. Vol. 3: Cognitive development. New York: Wiley.
Clark, H.H. (1973). Space. time. semantics, and the child. In T. Moore (Ed.). Cognitive development
und the ucyuisition of kunguuge. New York: Academic Press.
Corrigan. R., Halpern. E., Aviezer. 0.. & Goldblatt, A. (1981). The development of three spatial
concepts: In. on. under. Internutional Journal of Behaviorul Development. 3, 403-419.
Cromcr. R.F. (lY74). The dcvelopmcnt of language and cognition: The cognition hypothesis. In B.
Foss (Ed.). New perspectives in child development. Harmondsworth. Middlesex: Penguin.
DeLancey, S. (IYXS). The analysis-synthesis-lcxis cycle in Tiheto-Burman: A cast study in motivated
change. In J. Haiman (Ed.). Iconicity in syntux. Amsterdam: Benjamins.
Dromi, E. (lY87). Eurly lenicul development. Cambridge, UK: Cambridge University Press.
Farwell, C. (1977). The primacy of “goal” in the child’s description of motion and location. Papers
and Reports on Child Lunguuge Development. lh. 126-133.
Gentner. D. (lYX2). Why nouns are learned before verbs: Linguistic relativity versus natural
partitioning. In S.A. Kuczaj II (Ed.), Language development. Vol. 2: Lunguuge, thought. and
culture. Hillsdale. NJ: Erlbaum.
Gibson. E.J., & Spclke. E.S. (lY83). The development of perception. In J.H. Flavell & E.M.
Markman (Eds.), Mussen handbook o,f child psychology. Vof 3: Cognitrlze development. New
York: Wiley. Gopnik. A. (IYXO). The development of non-nominal expressions in 12-2~ month old children.
Unpublished doctoral dissertation. Oxford University.
Gopnik. A., & Choi. S. (IYYO). Do linguistic differences lead to cognitive differences? A cross-
linguistic study of semantic and cognitive development. First Language, IO, IVY-215.
Gopnik, A.. & Meltzoff, A. (lY86). Words, plans. things, and locations: Interaction between semantic
and cognitive development in the one-word stage. In S.A. Kuczaj II & M.D. Barrett (Eds.),
The dn~elopment of word meaning. Berlin: Springer-Verlag.
Greenfield. P.. & Smith. J. (1976). The .structure of communication in early languuge development.
New York: Academic Press.
Gruendel. J. (lY77a). Locative production in the single word utterance period: A study of up-down.
ore-off; and in-out. Paper presented at the Biennial Meeting of the Society for Research in
Child Development. New Orleans, 1977.
Gruendel, J. (lY77b). Referential extension in early language development. Child Devebpment. 4X.
1567-1576.
Halpern, E.. Corrigan, R.. & Aviezcr. 0. (IYXI). Two types of “under”‘.’ Implications for the
relationship between cognition and language. Internarionul Journul of Psyholinguistics. X-
3(24). 36-57.
Harris. M.. Barrett, M., Jones, D.. & Brookes. S. (198X). Linguistic input and early word meaning. Journul of Child Lunguuge. 1.7. 77-94.
Herskovits, A. (IYX6). Lunguuge and sputiul cognition: An interdisciplinary studv ofthe prepositions in
English. Cambridge, UK: Cambridge University Press.
Huttenlocher. J.. & Smiley. P. (lY87). Early word meanings: The case of object names. C‘ognirivr
P.q~cholog,v. 19. 63-89.
Huttcnlocher. J.. Smiley. P.. & Charney. R. (1983). Emergence of action categories in the child:
Evidence from verb meanings. Psychological Review. %I. 72-03.
120 S. Choi & M. Bowerman
Ingram, D. (1971). Transitivity in child language. Language, 47, 8888YOY.
Jackendoff, R. (1983). Semanrics and cognition. Cambridge, MA: MIT Press.
Jackendoff, R. (1990). Semanfic strucfurcs. Cambridge, MA: MIT Press.
Jackendoff. R.. & Landau, B. (1991). Spatial language and spatial cognition. In D.J. Napoli & J. Keg1 (Eds.). Bridges heiween psychology and linguistics: A Swarthmore Festschrifi for Lila Gkitman.
Hillsdale. NJ: Erlbaum.
Johnston. J., & Slobin. D.I. (1979). The development of locative expressions in English, Italian, Serbo-Croatian and Turkish. Journal of Child Language. 6, 547-562.
Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press.
Langacker. R.W. (1987). Foundations of cognirive grammar. Vol. 1: Theoretical prerequisites. Stan- ford. CA: Stanford University Press.
Landau, B., & Stecker, D.S. (in press). Objects and places: Syntactic geometric representations in early lexical learning. Cognitive Development.
Lee, H.B. (1989). Korean grammar. New York: Oxford University Press.
Leopold, W. (1939). Speech devrlopmenf of a bilingual child (Vol. 1). Evanston, IL: Northwestern University Press.
Levin, B. (1985). Lexical semantics in review: An introduction. In B. Levin (Ed.), Lexical semanfics
in review. Lexicon Project Working Papers, No. 1. Cambridge, MA: MIT Center for Cognitive
Science.
Levine, S.C.. & Carey. S. (1982). Up front: The acquisition of a concept and a word. Journal of Child Language. 9, 645-657.
Levinson, S.C. (1YYl). Relativity in spatial conception and description. Paper presented at Wenner-
Gren symposium “Rethinking linguistic relativity”, Ocho Rios, Jamaica. (To appear in J.J.
Gumperz & S.C. Levinson (Eds.), Rethinking linguistic relativity.)
Lindner. S. (1981). A lexico-semantic analysis of verb-particle constructions with up and our.
Unpublished doctoral dissertation, University of California at San Diego. Available from the
Indiana University Linguistics Club.
McCune-Nicolich. L. (1981). The cognitive bases of relational words in the single-word period.
Journul of Child Language. 8. 15-34.
McShane, J. (lY7Y). The development of naming. Linguistics, 17, 87YY905.
Miller. G.A.. & Johnson-Laird, P.N. (1976). Language and perception. Cambridge. MA: Belknap
Press of Harvard University Press.
Miller, W., bi Ervin. S. (1964). The development of grammar in child language. In U. Bellugi & R. Brown (Eds.), The acquisition of language. Monographs of the Society for Research in Child
Development, 29.02, Y -34.
Nelson, K. (1974). Concept. word, and sentence: Interrelations in acquisition and development.
Psychological Review. 81, 267-285.
Nelson, K., & Lucariello. J. (1985). The development of meaning in first words. In M.D. Barrett
(Ed.), Children’s single-word speech. New York: Wiley. Olson, D.R., & Bialystok. E. ( 1983). Spatial cognition: The structure and development of the mental
representation of spatial relations. Hillsdale. NJ: Erlbaum.
Piaget. J.. & Inhcldcr. B. (19S6). The child’s conception ofspace. London: Routledge & Kegan Paul.
Pinker, S. ( 1989). Learnability und cognilion: The acquisition of argument structure. Cambridge, MA:
MIT Press. Pye. C. (lY8.5). The acquisition of transitivity in Quiche Mayan. Papers and Reports on Child
Languugr Development. 24, 115~122. Pylyshyn. Z.W. (1977). Children’s internal descriptions. In J. Macnamara (Ed.). Language learning
and Ihought. New York: Academic Press.
Rispoli, M. (1987). The acquisition of the transitive and intransitive action verb categories in
Japanese. First Language. 7, 183-200.
Schlesinger, I.M. (lY77). The role of cognitive development and linguistic input in language
acquisition. Journal of Child Language. 4, 153-169. Slobin. D.I. (lY73). Cognitive prerequisites for the development of grammar. In C.A. Ferguson 8(
Learning to express motion events in English and Korean 121
D.I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston.
Slobin, D.I. (1985). Crosslinguistic evidence for the language-making capacity. In D.I. Slobin (Ed.).
The crosslinguistic study of language acquisition, Vol. 2: Theoretical issues. Hillsdale, NJ:
Erlbaum.
Slobin, D.I. (1991). Learning to think for speaking: Native language, cognition, and rhetorical style.
Paper presented at Wenner-Gren symposium ‘Rethinking linguistic relativity”, May, Ocho
Rios, Jamaica. (To appear in J.J. Gumperz & S.C. Levinson (Eds.). Rethinking linguistic relativily.)
Talmy. L. (1975). Semantics and syntax of motion. In J. Kimball (Ed.), Syntax and semantics. New York: Academic Press.
Talmy, L. (1983). How language structures space. In H. Pick & L. Acredolo (Eds.), Spatial orienfation: Theory, research, and application. New York: Plenum.
Talmy, L. (1985). Lexicalization patterns: Semantic structure in lexical forms. In T. Shopen (Ed.),
Language typology and syntactic description, Vol. III: Grammatical categories and the lexicon. Cambridge, UK: Cambridge University Press.
Tomasello. M. (1987). Learning to use prepositions: A case study. Journal of Child Language, 14, 79-98.
Van Geert, P. (1985/6). In, on, under: An essay on the modularity of infant spatial competence. Firsf Language, 6, 7-28.